GEOLOGICA BELGICA (2012) 15/1-2: 73-80

Impact of recent climate variability on an aquifer system in north Marc VAN CAMP, Kristine MARTENS & Kristine WALRAEVENS Laboratory for Applied Geology and Hydrogeology, University of Gent, Krijgslaan 281, 9000 Gent, Belgium

ABSTRACT. The last decade it has been realised that climate and global change can (and very likely will affect) groundwater reserves. This will have impact on both the economical (groundwater exploitation) and ecological (ecosystems) aspects of aquifer systems. A key issue is the intrinsic variability in hydrodynamics due to natural fluctuations of meteorological conditions. In this paper the results of a modelling study are presented that reconstructs the hydrodynamic evolution of the important Neogene aquifer system in North-East Belgium, between 1833 and 2005. Boundary conditions are defined on a monthly basis. The results show that besides the yearly seasonal fluctuations also multi-year to decadal variations occur. These are especially important in the topographic higher regions like on the Campine plateau and the Campine cuesta, and can severely affect local groundwater dependent ecosystems as found in some protected areas and nature reserves. The same cyclicity is recognized in the intensity and extent of seepage of deep Neogene groundwater in the Nete basin. KEYWORDS: modelling, hydrodynamics, Neogene, evolution

1. Introduction variation of meteorological parameters like precipitation and potential evapotranspiration, which are the main controlling physical entities Studying the impact of climate change on natural systems has become for recharge of aquifer systems. Without knowing this natural one of the main scientific challenges in recent years. Investigating variability it is very difficult to evaluate the impact of anthropogenic the influence of natural climate variability on natural systems isa influences like groundwater exploitation (pumping) or the climatic prerequisite to the understanding of system responses to external shift expected to occur by global warming. This natural variability has forcing stresses and helps to quantify the variability in resources. been investigated for the Neogene aquifer by means of a numerical As groundwater dependent ecosystems largely depend on the groundwater model and a simulation of the hydrodynamic history availability of groundwater, they are very sensitive to groundwater during more than one and a half century, only taking into account changes and variations. These can be anthropogenic, as induced by natural conditions, excluding anthropogenic factors like pumping groundwater pumping, which is often local. Fluctuations on a larger or changes in land use and hydrography. Spatial variation in aquifer spatial scale may be caused by meteorological changes. Sequences recharge was also not considered. of multiple dry or wet years can trigger periods with overall lower or higher groundwater levels. To quantify the potential risk of hydro- 2. Description of the Neogene aquifer system ecosystems, understanding the hydrodynamic response of the aquifer to these extreme meteorological conditions is useful. 2.1. Geological and hydrostratigraphical setting The northern parts of the provinces of Antwerp and , in The aquifer’s substratum is formed by the Boom Clay (Fig 2), which the north-east of Belgium, comprise a wealth of small streams and is dipping in the northern and north-eastern direction. It is covered rivulets in the extensive Nete Basin (Fig 1), where many valuable by Paleocene and Miocene sandy deposits from the Formations of groundwater dependent ecosystems have developed. Also the more Eigenbilzen, Voort, Berchem, Bolderberg, Diest and Kattendijk. elevated parts of the region (the Campine Plateau in the south-east These are hydraulically well connected and form a single aquifer, and the Campine Cuesta in the north) contain many nature reserves which is identified with HCOV(“Hydrogeologische Codering van and ecological valuable areas. de Ondergrond van Vlaanderen”, Meyus et al, 2000) code 250. The To secure the future of these valuable groundwater dependent Miocene aquifer is covered by more clayey layers which belong ecosystems, an understanding of the intrinsic variations of to the Pliocene Formation of Lillo. This aquitard is identified with groundwater levels and hydrodynamics in general is necessary. the HCOV code 240 and separates the underlying Miocene aquifer A first step is to quantify the changes that originate from natural from the overlying Pleistocene and Pliocene aquifer (HCOV 0230). 0 0 0 0 4 2 The Netherlands 0 0 0 0 3 2

0 Model boundary 0

0 Campine Cuesta 0 2 2 0 0 0 0 1 2 0 0 0

0 Nete Basin 0 2 Campine 0 0 0 0

9 Plateau 1 Boom Clay outcrop 0 0 0 0 8 1

Figure 1. Location and 0 0

0 physiography of the Neogene 0 7

1300001 140000 150000 160000 170000 180000 190000 200000 210000 220000 230000 240000 aquifer. 74 M. Van Camp , K. Martens & K. Walraevens Figure 2. Cross section. RECHARGE

groundwater divide 100

F DISCHARGE RECHARGE DISCHARGE of Kas ) terle l 50 e s

a groundwater divide

m ground level ( F of Mol

n Formation

o Formation i 0 Campine Complex t of Diest Fo

a of Diest r mat v ion o e f l Bra e ssch F aat c

i -50 o r Formatio h m Fo n of rmatio Lillo p a n F t of Kattendijk a Boom o io r r Formation of Diest F m n g Clay of Eig a o o enb f -100 ilze ti B p n o n o o ld o T e Boom f r V b o e Clay o rg r -150 t Formation of Berchem LEGEND Formation Boom Main groundwater flow cycle of Tongeren F in Neogene aquifer o o f r Clay -200 T m Formation of Diest o at ng io Boom Clay Lithostratigrapic Unit

e n fault fault fault fault r fault en 0 5 101520253035404550556065707580859095100 Distance along profile (km)

This latter includes the late Pliocene and Pleistocene sands of the 2.3. Groundwater exploitation Formations of Lillo, , Poederlee, Mol, Merksplas and Brasschaat. In most of the study region this aquifer is phreatic as The Neogene aquifer is the most important aquifer system in it is only covered by younger Quaternary sands. In the north of the north-east Belgium. Due to its considerable thickness, increasing province Antwerp a very heterogeneous complex is deposited on top to the north, and high hydraulic conductivity and accordingly high 2 of the Pleistocene and Pliocene aquifer, making it semi-confined. This transmissivity (up to a few thousand m /day), it can provide high complex (HCOV code 220) contains sand, silt, clay and peat layers amounts of water even with relatively low drawdowns (often no more and is both vertically as laterally very heterogeneous with strongly than a few meters). Over the last decades numerous pumping sites varying textures. have been installed, ranging in size and capacity from local home users to drinking water companies with well fields capturing up to 10 3 2.2. Topography and hydrography million m /year. In recent years total exploitation rate in the Neogene aquifer is around 150 Mm3/year (Coetsiers, 2007), which is nearly 12% of the estimated total aquifer recharge. Most of the water is taken Topography (Fig 1) drops from the east to the west and to the north. from the Formation of Diest and in the northern region also from the The highest levels are found on the Campine Plateau (around + 70 Formation of Brasschaat, often found at 30 or 40 m depth. masl) in the southeast, while the lowest levels occur in the lower Nete valley in the west (less than + 10 masl). In the north a west- 2.4. Piezometric levels and hydrodynamics east oriented ridge separates the Nete river valley from low lying polders in The Netherlands. This cuesta is formed by clay layers in Measurements in more than 500 wells were used to make a piezometric the Campine Formation. The southern side of the cuesta is steeper map (Fig 3) of the Neogene aquifer for the year 2003 (Coetsiers, 2007). than the northern side. Although most mesurements were not taken at exactly the same time, 0 0 0 0 4 2 0 0 0 0

3 Kalmthout 2 0 0 0 0 2 2 0 0 0 0 1 2 Hechtel 0

0 0 0 0 2 0

0 LEGEND 0 0

9 30 piezometric level (m asl) 1

piezometric measurement 0

0 selected observed series 0 0

8 model boundary 1

Figure 3. Piezometric map anno 2003. 140000 150000 160000 170000 180000 190000 200000 210000 220000 230000 Impact of recent climate variability on an aquifer system in north Belgium 75 but were rather spread over the year 2003, the resulting map gives a characterized by an alternation of periods with lower than normal representative view of the piezometric surface. It captures the main levels, like 1990-1992 and 1996-1998, and higher levels, like 1987- hydrodynamic trends and groundwater flow cycles in the region. The 1989 and 1999-2002. Winter levels in the dry periods can be lower highest piezometric levels are found in the southeast on the Campine than summer levels in the wet periods. The same pattern is found Plateau, where ground level reaches around 70 masl. Piezometric in observation wells located close to the groundwater divide on the levels decrease towards the east, the west and the north. Topography Campine Cuesta, in the norhwestern part of the region, like the well in is the controlling factor for the main groundwater cycle. On the Kalmthout (Fig 4b). Here, summer minima show a range of up to two Campine Plateau itself the hydrographic network is sparse and most meters, which is more than the average difference between winter and rainwater infiltrates. From here groundwater flows mainly laterally summer. The general fluctuations in this longer series (from 1981 on) towards the low lying Nete Valley, where seepage drain it into the Nete correspond very well with the behaviour of the well in Hechtel. The river and contributaries. As the deep Neogene groundwater usually range of the seasonal variations depends on the specific yield of the has high dissolved iron concentrations because of the reducing redox phreatic layer, which is related to the lithology of the subsurface and conditions, brown coloured precipitation in many ditches, caused by can be different for these two sites. In contrast, a typical well located oxidation of the upwelling water, indicates the occurrence of these in the Nete Valley in Westerlo (Fig 4c), shows much less pronounced seepage cycles in the valleys. Under the northern region, close to the interyearly variations. Although high winter levels can still vary border with the Netherlands, a secondary groundwater divide is found considerably from year to year, depending on winter rainfall, the low at the topographic height of the Campine Cuesta. summer levels range much less, most of them lie between 15.2 and A whole set of observation wells is being used now by the Flemish 15.6 m. The main reason is the density of the stream network and the government (department VMM) to monitor groundwater levels, but proximity and good hydraulic contact of surface water that limits the most of these wells were installed only during the last decade. Few lowering of the water table. long time series, spanning multiple decades, exist. The oldest series started at the end of the 1970s. Some typical data sets are given in Fig 2.5. Aquifer recharge 4. The location of the series are indicated on Fig 3. A first well (Fig 4a) is located in the south-east on the Campine Plateau in Hechtel. The main source of water in the aquifer system is recharge from Levels here are very high (around 56 m asl) because of the high precipitation. Aquifer recharge is estimated using data series of the topography, and in the series the regular seasonal fluctuations (high Ukkel meteorological station, obtained from the global historical winter and low summer levels) can clearly be seen. But the most climatology network (GHCN). Although Ukkel lies outside the important characteristic is the presence of a multi-year trend that is modelled region (around 30 km from the southwest border), it is by far

57 ) W A T m ( l e v e l 56 c i r t e m o z e i P

55 1980 1990 2000 2010 TIME (year) 23 ) W A

T 22 m ( l e v e l 21 c i r t e m o

z 20 e i P

19 1980 1990 2000 2010 TIME (year)

17 ) W A T m ( l e v e l 16 c i r t e m o z e i P

15 Figure 4. Selected piezometric time 1980 1990 2000 2010 series. TIME (year) 76 M. Van Camp , K. Martens & K. Walraevens

Precipitation ( ) Figure 5. Result of the soil moisture water balance model for the period 1996-2005. 250 200 150 100 50 0 Monthlyprecipitation(mm) 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 ) ) h t

m soil moisture content ( ) and aquifer recharge ( ) n o m

( 100 160 m / t Field Capacity = 100 mm n m

e 80 t

120 m ( n o 60 e C g

80 r e a r

40 h u t c s e i 40 r

o 20 r M e f l i i 0 0 u o q S

1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 A Temperature ( ), PET( ) and AET ( ) 25 160 20 120 15 10 80 5

Temperature 40 0 -5 0 PETand AET (mm/month) 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 the longest series available in Belgium, and thus allowed for a much for the last decade of the simulation period (1996-2005) are presented longer simulation period. Recharge is computed with a soil moisture graphically as an example (Fig 5). In the upper graph the measured balance (SMB) approach as introduced by Thornthwaite and Mather monthly rainfall is given. The middle graph shows the calculated (1955) and applied by Mather (1978 and 1979), using precipitation aquifer recharge (in mm/month) as a bar chart and the soil moisture and temperature data. Monthly PET (potential evapotranspiration content (hatched graph). The maximal value at field capacity was in mm/day) is estimated from temperature using the Thornthwaite estimated at 100 mm. The lower graph shows measured monthly (1948) equation: averaged temperatures and the values of monthly potential (PET) and a ⎛ Li ⎞⎛ Ni ⎞⎛10Ta ⎞ actual (AET) evapotranspiration calculated with the meteorological PET i =16⎜ ⎟⎜ ⎟⎜ ⎟ ⎝12 ⎠⎝ 30 ⎠⎝ I ⎠ data of Ukkel. In Fig 5 it can easily be recognized that aquifer recharge with: T is the average daily temperature of the month being calculated is limited to the winter months. The main controlling factor is that a PET and AET are very low in wintertime while they usually exceed (°C), Ni is the number of days in the month I being calculated (-), Li is the average day length of the month I being calculated (hours). precipitation in summer months. Intermonthly variation of rainfall is small but interyearly variation is large, while intermonthly variation α = (6.75 10-7) I3 – (7.71 10-5 ) I2 + (1.79 10-2) I + 0.49 in PET and AET values is large, but interannual variation is small. I = heat index defined as (-): Average recharge rate over the whole period 1833-2005 is 239

1.514 mm/year, with an average precipitation of 790 mm/year. Interyearly 12 ⎛T ⎞ I = ∑ ⎜ ai ⎟ variation can be seen in Fig 6. A longer time cyclicity can be i =1 ⎝ 5 ⎠ recognized in the 10 year simple running average, which is defined as This approach has the benefit that it requires only a single the simple arithmetic average of the last ten values (including the year meteorological parameter: temperature. The calculations for the itself). Drier decades can be found around 1860-1870, 1900-1910 soil moisture balance are done based on monthly totals (for rainfall) and 1950-1960. They apparently occur with a periodicity of 40 to 50 and averages (for temperature). The program WATBUG (Willmott, years. Wetter periods are interspersed with the wettest decade recently 1977) was used for this purpose. Results of the SMB model include between 1980 and 1990. In the dry periods average recharge is a bit potential evapotranspiration, actual evapotranspiration, deficit, soil less than 200 mm/year, while the wet decades have average rates close water content and aquifer recharge. The results of the SMB model to 300 mm, around 50% higher than de dry period average value.

600 Yearly aquifer recharge 10 year simple moving average 500 Linear trend ) r a e y / 400 m m ( e g r

a 300 h c e r r e f

i 200 u q A

100

Figure 6. Calculated yearly aquifer 0 recharge 1833-2005, 10 year simple 1830 1840 1850 1860 1870 1880 1890 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 2010 moving average and linear trendline. TIME (year) Impact of recent climate variability on an aquifer system in north Belgium 77 A linear correlation analysis shows a slight increasing trend with an boundaries. The south border is located at the northern outcrop of increase of ca 3.8 mm a decade, but the coefficient of determination the Boom Clay. Inflow is considered here as negligible. The eastern (r squared) is only 0.04. border is chosen along a flowline and is therefore by definition a no flow boundary. Along the western border, groundwater is drained 3. Groundwater flow model towards the river, which is included as a river. Baseflow to the river system has been included in the model with a diffuse approach. The groundwater flow model used for the simulations is based on the In the Nete Valley the number of rivers, streams and small ditches model conceptualized and implemented by Coetsiers (2007). It uses is too large to consider them as individual river entities. Instead in the MODFLOW simulator (Harbaugh and Mc Donald , 1996). The each model cell a drainage level is defined, which corresponds with model domain is discretised using a rectangular grid with a uniform the average bottom level that the river reaches. This bottom level is spacing of 500 m. The whole modelled region includes an area of derived from a DTM, not from in situ field measurements. In each 100 km by 75 km, subdivided in 200 columns and 150 rows. The model cell, also outside the Nete valley, the possible drainage is south border (Fig 1) lies at the outcrop of the underlying Boom Clay defined using the MODFLOW DRAIN module. If calculated water (Formation of ), which is considered here as the basement of the table is lower than the predefined drain level, the discharge flux is model. Flow through this basement is considered negligible compared zero: to the fluxes in the overlying aquifer. The south border is therefore a WT <= LDRN → QDRN = 0 no flow boundary. The west boundary of the model is located at the With WT = calculated water table level in a cell, LDRN = predefined Scheldt river. As the Neogene aquifer is phreatic there, the river can drainage level in a cell, and QDRN = amount of water drained from the be considered as a hydraulic boundary. The river is incorporated in cell. the model with constants head cells. The east model border lies at If the water table is above the defined drainage level, the discharge or near to a flowline as can be seen on the piezometric map (Fig 3). flux is proportional to the head difference: This border is treated as a no flow boundary. The north border lies in The Netherlands and its position is chosen rather arbitrarily. At this WT > LDRN → QDRN = CF*(WT-LDRN) border constant heads were entered into the model. The levels were The constant CF (units: m2/day) defines the hydraulic contact between estimated based on the topographic elevations, derived from a digital the aquifer and the drain mechanism. terrain model (DTM). Hydraulic properties are based on results of By combination of the recharge and drainage modules of the pumping tests that were done in the past in some of the exploitation MODFLOW model, the extension of the recharge and discharge areas well fields. Average hydraulic conductivity for the lower part of the can be calculated. From the model’s output file, the discharge flux of Neogene aquifer is 10.1 m/d, for the upper part 15.3 m/d. The cover each model cell in the phreatic layer can be retrieved. If the discharge sands have an average value of 12.9 m/d (Meyus et al., 2004) rate is zero, only recharge takes places in this cell and it belongs to the The main groundwater source is recharge from rainfall. The “recharge” area. If there exists a discharge flux, the model cell belongs model uses a uniform distribution of aquifer recharge, with a long to the “discharge” area, although it should be realised that also in these term average of 239 mm/year. It should be kept in mind that spatial cells rainwater is entering the aquifer system. By accounting for the variations in recharge pattern will occur, related to differences in total extent of the discharge area, where groundwater discharges into soil texture, land use and vegetation, but these are smoothed by the the river network, total and average seepage fluxes can be calculated. use of a single uniform value. First a steady state run with the long First a steady state run with exploitation data and boundary term average recharge of 239 mm/year was done. The calculated conditions for 2003 was performed (Fig 7), to check the ability of piezometric levels were used as initial heads for a transient simulation the model to reproduce the recently measured piezometric distribution starting in 1833 with monthly stress periods. The first years of this (Fig 3). The correspondence of this calibration was sufficient to simulation has to seen as a spinup period.In the transient simulation accept the entered parameterisation. Model calculated baseflow in the run, recharge is defined with monthly stress periods. These were subbasin of the river “Kleine Nete” was compared with estimation calculated with the soil moisture balance model (see 2.5) of baseflow derived from measured river flow rates at a monitoring The main groundwater sink mechanisms are groundwater station in (Fig 8) (HIC, 2004) . As an approximation, discharge into the river system and groundwater flow over the outer the minimum daily flowrate in each month was considered as model boundaries. Groundwater exploitation is not considered in constituted largely of baseflow alone. Deviations between modelled the simulation 1833-2005 as the objective was to investigate natural and measured values for the year 2003 were on the order of ca 10%, variations in the aquifer hydrodynamics. Cross border flow is only which is probably well within the limits of accuracy of the simple relevant over the north border as the other borders are no flow approximation of river baseflow.

Figure 7. Results of the calibration run. 78 M. Van Camp , K. Martens & K. Walraevens 40 Figure 8. Monthly minimum, maximum and average daily flow rate on the Kleine Nete river in Grobbendonk.

30 ) c e s / 3 m ( E T A

R 20 E G R A H C S I D

10

0 2003 2004 2005 2006 2007 2008 TIME (years)

The final model run uses a transient flow regime with stress period The long term evolution of piezometric levels can be illustrated length of one month. Each stress period is divided into 5 time steps. by time series graphs of the calculated hydraulic head at two selected Initial heads for the run were obtained from the steady state model run locations (Fig 9). The levels are plotted as deviations (in m) from the with the average recharge value of 239 mm/year. average of the total series. The main trend is indicated by the two year running average which attenuates seasonal fluctuations. In the upper 4. Simulation results graph the evolution on the Campine Plateau is given. Here the main variations in level occur on a multi-year time scale and the yearly The 1833-2005 model run simulates the situation without groundwater seasonal fluctuations are less important than the longer cyclicity. The extraction, representing the natural, pristine state of the aquifer. Output three main dry periods of 1860-1870, 1900-1910 and 1950-1960 are of the model can be represented in the form of graphs and maps. easily identifiable with levels that are 1 to 2 meters below average. Results indicate that, besides the yearly seasonal variation in the main Since the early 1980s, levels were almost always above the long term groundwater flow cycle, also a longer multi-year cyclicity occurs. average of the whole series. The lower graph shows a representative This is mainly in the regions where the aquifer is being recharged. location in the Nete valley basin. In this series, variation can be largely Both simulated and measured series of the last two or three decades attributed to the seasonal cycle exclusively. show this behaviour. The model outputs monthly distributions of As can be seen on the time graphs, the maximum range of hydraulic heads, fluxes from or to defined boundary conditions and variation over the years is a few meters, between 3 and 4 meters on flow rates inside the model domain (between model cells and layers). the Campine Plateau. This is only a small fraction of the total range in

2 )

m Monthly values (

e 2 year running average g a

r 1 e v a - l e v 0 e l c i r t e m

o -1 z o e i P -2

1830 1840 1850 1860 1870 1880 1890 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 2010 TIME (years) 2 Monthly values 2 year running average 1

0

-1

Figure 9. Calculated evolution of (m) average - level Pieozometric piezometric levels at 2 selected locations: -2 on the Campine plateau (upper graph) and in 1830 1840 1850 1860 1870 1880 1890 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 2010 the Nete basin (lower graph). TIME (years) Impact of recent climate variability on an aquifer system in north Belgium 79 Figure 10. Difference in piezometric level 0

of the winter 2002 maximum compared to 0 0 0

the long term average levels. 4 2 0 0 0 0 3 2 0 0 0 0 2 2 0 0 0 0 1 2 3.4 3.2

0 3 0

0 2.8 0

0 2.6 2 2.4 2.2

0 2 0

0 1.8 0

9 1.6 1 1.4 1.2

0 1 0

0 0.8 0

8 0.6 LEGEND 1 0.4 MODEL BOUNDARY 0.2 0 RIVER

140000 150000 160000 170000 180000 190000 200000 210000 220000 230000 piezometric levels in the whole aquifer system, as there is at least 50 same conclusion is valid for the extension of the discharge region; m difference beween the Campine Plateau and the Scheldt river in the in summer time discharge has the smallest extension, between west. Consequently, the general flow pattern will not fundamentally 1000 and 1500 km2, and is limited to the main river valleys. Many change between seasons or between extremely dry and wet years. small ditches can dry up in a dry summer period. In winter time, Deviations from the long term average can be greater on the Campine the discharge region’s size can triple to more than 3000 km2. The Plateau and Campine Cuesta as can be seen on the map of Fig 10. For average discharge flux intensity can be calculated by dividing the the wet winter situation of January 2002, the piezometric levels are total baseflow for each month by the extension over which it occurs. compared with the average levels of the whole simulated period. They A cross-plot of the baseflow and the average discharge intensity as can be as much as 3 m above normal. function of the extension of the discharge region (Fig 12) shows that the relation between total baseflow and extension of the seepage As a result of the variations in hydraulic heads, the intensity of the region is non-linear, but rather exponential. This means that, as its groundwater cycles will vary over time. This is especially important areal size increases, also the intensity will increase. In dry summer for the upward groundwater flow in the Nete Valley, where deep months, average seepage flux is around 1 mm/day, but it increases Neogene groundwater is upwelling and is drained away by the river to 2.5 mm/day in wet winter months. The exponential fits indicated network. Here both size of the region as the intensity of the discharge in Fig 12 have a coefficient of determination (r squared) of 0.89 for flow will vary with time (Fig 11). Total calculated discharge into the the relationship extension-discharge intensity and 0.98 for the relation river shows a strong seasonal behavior with low summer values, extension-baseflow. nearly always between 15 and 20 Mm3/month. There is however a rather small interyearly variation on the summer minimum baseflow. 5. Conclusions Maximum baseflow appears in wintertime, but varies from year to year over a wide range between around 40 and 140 Mm3/month. The natural variations of the main groundwater cycle characteristics Interannual variability is much larger than for the summer lows. The in the Neogene aquifer system in north-east Belgium have been

4000 ) 2 m k (

a 3000 r e a e g r

a 2000 h c s i d t

n 1000 e t

x Monthly values E 2 year running average 1600

140 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 2010 /month) 3 120

100

80

60

40

20

Total monthly baseflow (Mm 0 Figure 11. Evolution of the monthly 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 2010 baseflow and the extent of the discharge TIME (years) area. 80 M. Van Camp , K. Martens & K. Walraevens

4 160 Figure 12. Relation between the model calculated monthly baseflow and the extent of the discharge area. Average discharge intensity (mm/day) Total monthly baseflow

3 120 ) h ) t y n a o d / m / 3 m m m ( M x ( u l w f o l e f g e r

2 80 s a a h b c s y i l d h t e n g o a r m l e v a t A o T 1 40

0 0 0 1000 2000 3000 4000 Extent discharge area (km2) investigated using a numerical model. Using the longest available Mather, J.R., 1979. Use of the climatic water budget to estimate streamflow. meteorological series in Belgium, from the Ukkel station, a transient In: Mather, J.R. (ed.) Use of the climatic water budget in selected flow simulation of a 173 year long period from 1833 till 2005 was environmental water problems, Elmer, N.J., C.W. Thornthwaite performed. Aquifer recharge is calculated on a monthly basis with Associates, Laboratory of Climatology, Publications in Climatology, 32, a soil moisture balance model and shows the presence of three drier (1), 1–52. Meyus, Y., Batelaan, O. & De Smedt, F., 2000. Concept Vlaams Grondwater decades which occur with intervals of 40 to 50 years: in 1850-1860, Model (VGM), technisch concept van het VGM, Deelrapport I: in 1900-1910 and in 1950-1960. The model simulation uses monthly Hydrogeologische Codering van de Ondergrond van Vlaanderen (HCOV). stress periods with changing but spatially uniform recharge rates (in Dutch) and uses a combination of the recharge and drainage capabilities of Meyus,Y., Adyns, Severyns, J., Batelaan, O. & De Smedt, F., 2004. the simulator to calculate extension and intensity of the discharge Ontwikkeling van regionale modellen ten behoeve van het Vlaams component of the main groundwater cycle. The model results indicate Grondwater Model (VGM) in GMS/MODFLOW Perceel 1: Het Centraal the existence of a multi-year cyclicity in the piezometric levels which Kempisch model Deelrapport 1: Basisgegevens en conceptueel model. is more pronounced in the topographic higher areas where most of Vrije Universiteit Brussel . Vakgroep Hydrologie en Waterbouwkunde (in the aquifer recharge takes place. The multi-year variations are here Dutch) more important than the yearly seasonal winter-summer cycle. Thornthwaite, C.W., 1948. An approach toward a rational classification of Unfortunately, no very long observation series of piezometric levels climate: Geographical Review, 38, 55–94. exist, to compare with model results. The oldest series started back Thornthwaite, C.W. & Mather, J.R., 1955. The Water Balance. Publications in Climatology 8,1, Centerton, New Jersey. near the end of the 1970s. Discharge of deep Neogene groundwater in Willmott, C.J., 1977. Watbug: a Fortran IV Algorithm for Calculating the the Nete Valley shows a strong seasonal dependency for intensity and Climatic Water Budget, Publications in Climatology 30, 2. Centerton, areal extent, and the interannual variation of peak winter maxima is New Jersey. much larger than for summer lows which are more constant.

6. References

Coetsiers, M. & Walraevens, K., 2006. Chemical characterisation of the Neogene Aquifer, Belgium. Hydrogeology Journal 14(8), 1556-1568. Coetsiers, M., 2007. Investigation of the hydrogeological and hydrochemical state of the Neogene Aquifer in Flanders with use of modeling and istopes hydrochemistry. PhD thesis, Ghent University, Belgium, 237 pp. Coetsiers, M., Van Camp, M. & Walraevens, K., 2003. Reference Aquifer: The Neogene Aquifer of Flanders. In: Edmunds, W.M. and Shand, P. (eds.), Natural Baseline Quality in European Aquifers: A Basis for Aquifer Management. EC-project EVK1-CT-1999-00006. Final Report, British Geological Survey, Wallingford Harbaugh, A.W. & Mcdonald, M.G., 1996. User’s documentation for MODFLOW-96, an update to the U.S. Geological Survey modular finite- difference ground-water flow model: U.S. Geological Survey Open-File Report 96-485, 56 p. Hic, 2004. Hydrologisch Jaarboek 2003. Waterbouwkundig Laboratorium en Hydrologisch Onderzoek, Borgerhout, Belgium. 137 p. Mather, J.R., 1978. The climatic water balance in environmental analysis: Manuscript received 08.02.2010, accepted in revised form 31.8.2011, Lexington, Mass., D.C. Heath and Company, 239 p. available on line 10.12.2011.